double scale_x,
double scale_y )
{
- int i, j;
- int x;
- int x_step = ( 1 << SCALE_SHIFT ) / scale_x;
- int y_step = ( 1 << SCALE_SHIFT ) / scale_y;
+ register int i, j;
+ register int x_step = ( 1 << SCALE_SHIFT ) / scale_x;
+ register int y_step = ( 1 << SCALE_SHIFT ) / scale_y;
+ register int x, x_scaled;
for ( i = 0; i < ( render_y1 - render_y0 ); i++ )
{
- const guchar *src = src_buf + ( ( ( i + render_y0 ) * y_step + y_step / 2 ) >> SCALE_SHIFT ) * src_rowstride;
+ const guchar *src = src_buf + ( ( ( i + render_y0 ) * y_step + ( y_step >> 1 ) ) >> SCALE_SHIFT ) * src_rowstride;
guchar *dest = dest_buf + i * dest_rowstride;
-
- x = render_x0 * x_step + x_step / 2;
-
+ x = render_x0 * x_step + ( x_step >> 1 );
+
for ( j = 0; j < ( render_x1 - render_x0 ); j++ )
{
- const guchar *p = src + ( x >> SCALE_SHIFT ) * 4;
- guint32 *p32;
-
- p32 = ( guint32 * ) dest;
- *p32 = *( ( guint32 * ) p );
-
- dest += 4;
+ x_scaled = x >> SCALE_SHIFT;
+ *dest++ = src[ x_scaled << 1 ];
+ *dest++ = src[ ( ( x_scaled >> 1 ) << 2 ) + ( ( j & 1 ) << 1 ) + 1 ];
x += x_step;
}
}
guchar **src,
int x_init, int x_step, int src_width )
{
- int x = x_init;
- int i, j;
+ register int x = x_init;
+ register int i, j, x_scaled, y_index, uv_index;
while ( dest < dest_end )
{
- int x_scaled = x >> SCALE_SHIFT;
- int *pixel_weights;
-
- pixel_weights = weights + ( ( x >> ( SCALE_SHIFT - SUBSAMPLE_BITS ) ) & SUBSAMPLE_MASK ) * n_x * n_y;
+ unsigned int y = 0, uv = 0;
+ int *pixel_weights = weights + ( ( x >> ( SCALE_SHIFT - SUBSAMPLE_BITS ) ) & SUBSAMPLE_MASK ) * n_x * n_y;
- unsigned int y1 = 0, cb = 0, y2 = 0, cr = 0;
+ x_scaled = x >> SCALE_SHIFT;
+ y_index = x_scaled << 1;
+ uv_index = ( ( x_scaled >> 1 ) << 2 ) + ( ( dest_x & 1 ) << 1 ) + 1;
+
for ( i = 0; i < n_y; i++ )
{
- guchar *q = src[ i ] + x_scaled * 4;
int *line_weights = pixel_weights + n_x * i;
+ guchar *q = src[ i ];
- for ( j = 0; j < n_x; j++ )
+ for ( j = 0; j < n_x; j ++ )
{
unsigned int ta = line_weights[ j ];
- y1 += ta * q[ 0 ];
- cb += ta * q[ 1 ];
- y2 += ta * q[ 2 ];
- cr += ta * q[ 3 ];
-
- q += 4;
+ y += ta * q[ y_index ];
+ uv += ta * q[ uv_index ];
}
}
- dest[ 0 ] = ( y1 + 0xffff ) >> 16;
- dest[ 1 ] = ( cb + 0xffff ) >> 16;
- dest[ 2 ] = ( y2 + 0xffff ) >> 16;
- dest[ 3 ] = ( cr + 0xffff ) >> 16;
-
- dest += 4;
+ *dest++ = ( y + 0xffff ) >> SCALE_SHIFT;
+ *dest++ = ( uv + 0xffff ) >> SCALE_SHIFT;
x += x_step;
+ dest_x++;
}
return dest;
guchar **src,
int x_init, int x_step, int src_width )
{
- int x = x_init;
- guchar *src0 = src[ 0 ];
- guchar *src1 = src[ 1 ];
+ register int x = x_init;
+ register guchar *src0 = src[ 0 ];
+ register guchar *src1 = src[ 1 ];
+ register unsigned int p;
+ register guchar *q0, *q1;
+ register int w1, w2, w3, w4;
+ register int x_scaled, x_aligned, uv_index;
while ( dest < dest_end )
{
- unsigned int y1, cb, y2, cr;
- int x_scaled = x >> SCALE_SHIFT;
- int *pixel_weights;
- guchar *q0, *q1;
- int w1, w2, w3, w4;
-
- q0 = src0 + x_scaled * 4;
- q1 = src1 + x_scaled * 4;
-
- pixel_weights = weights + ( ( x >> ( SCALE_SHIFT - SUBSAMPLE_BITS ) ) & SUBSAMPLE_MASK ) * 4;
+ int *pixel_weights = weights + ( ( x >> ( SCALE_SHIFT - SUBSAMPLE_BITS ) ) & SUBSAMPLE_MASK ) * 4;
+
+ x_scaled = x >> SCALE_SHIFT;
w1 = pixel_weights[ 0 ];
w2 = pixel_weights[ 1 ];
w3 = pixel_weights[ 2 ];
w4 = pixel_weights[ 3 ];
- y1 = w1 * q0[ 0 ];
- cb = w1 * q0[ 1 ];
- y2 = w1 * q0[ 2 ];
- cr = w1 * q0[ 3 ];
-
- y1 += w2 * q0[ 4 ];
- cb += w2 * q0[ 5 ];
- y2 += w2 * q0[ 6 ];
- cr += w2 * q0[ 7 ];
-
- y1 += w3 * q1[ 0 ];
- cb += w3 * q1[ 1 ];
- y2 += w3 * q1[ 2 ];
- cr += w3 * q1[ 3 ];
+ /* process Y */
+ q0 = src0 + ( x_scaled << 1 );
+ q1 = src1 + ( x_scaled << 1 );
+ p = w1 * q0[ 0 ];
+ p += w2 * q0[ 2 ];
+ p += w3 * q1[ 0 ];
+ p += w4 * q1[ 2 ];
+ *dest++ = ( p + 0x8000 ) >> SCALE_SHIFT;
+
+ /* process U/V */
+ x_aligned = ( ( x_scaled >> 1 ) << 2 );
+ q0 = src0 + x_aligned;
+ q1 = src1 + x_aligned;
+ uv_index = ( ( dest_x & 1 ) << 1 ) + 1;
+ p = w1 * q0[ uv_index ];
+ p += w2 * q0[ uv_index ];
+ p += w3 * q1[ uv_index ];
+ p += w4 * q1[ uv_index ];
+ *dest++ = ( p + 0x8000 ) >> SCALE_SHIFT;
- y1 += w4 * q1[ 4 ];
- cb += w4 * q1[ 5 ];
- y2 += w4 * q1[ 6 ];
- cr += w4 * q1[ 7 ];
-
- dest[ 0 ] = ( y1 + 0x8000 ) >> 16;
- dest[ 1 ] = ( cb + 0x8000 ) >> 16;
- dest[ 2 ] = ( y2 + 0x8000 ) >> 16;
- dest[ 3 ] = ( cr + 0x8000 ) >> 16;
-
- dest += 4;
x += x_step;
+ dest_x++;
}
return dest;
guchar **src, int src_channels,
int x_start, int src_width )
{
- unsigned int y1 = 0, cb = 0, y2 = 0, cr = 0;
- int i, j;
+ register unsigned int y = 0, uv = 0;
+ register int i, j;
+ int uv_index = ( ( dest_x & 1 ) << 1 ) + 1;
for ( i = 0; i < n_y; i++ )
{
for ( j = 0; j < n_x; j++ )
{
- unsigned int ta;
- guchar *q;
+ unsigned int ta = 0xff * line_weights[ j ];
if ( x_start + j < 0 )
- q = src[ i ];
+ {
+ y += ta * src[ i ][ 0 ];
+ uv += ta * src[ i ][ uv_index ];
+ }
else if ( x_start + j < src_width )
- q = src[ i ] + ( x_start + j ) * src_channels;
+ {
+ y += ta * src[ i ][ ( x_start + j ) << 1 ];
+ uv += ta * src[ i ][ ( ( ( x_start + j ) >> 1 ) << 2) + uv_index ];
+ }
else
- q = src[ i ] + ( src_width - 1 ) * src_channels;
-
- ta = 0xff * line_weights[ j ];
-
- y1 += ta * q[ 0 ];
- cb += ta * q[ 1 ];
- y2 += ta * q[ 2 ];
- cr += ta * q[ 3 ];
+ {
+ y += ta * src[ i ][ ( src_width - 1 ) << 1 ];
+ uv += ta * src[ i ][ ( ( ( src_width - 1 ) >> 1 ) << 2) + uv_index ];
+ }
}
}
- dest[ 0 ] = ( y1 + 0xffffff ) >> 24;
- dest[ 1 ] = ( cb + 0xffffff ) >> 24;
- dest[ 2 ] = ( y2 + 0xffffff ) >> 24;
- dest[ 3 ] = ( cr + 0xffffff ) >> 24;
+ *dest++ = ( y + 0xffffff ) >> 24;
+ *dest++ = ( uv + 0xffffff ) >> 24;
}
dest_rowstride,
src_buf, src_width, src_height, src_rowstride,
scale_x, scale_y );
- return ;
+ return;
}
filter.overall_alpha = 1.0;
make_weights ( &filter, interp_type, scale_x, scale_y );
-fprintf( stderr, "RESCALE: %d %d\n", filter.x.n, filter.y.n );
if ( filter.x.n == 2 && filter.y.n == 2 )
{
#ifdef USE_MMX
projects / melted / blobdiff